Under the shadow of the trillion-yuan Zhipu, Chinese internet giants face an "identity crisis"

On June 22, the market value of Zhipu AI exceeded HK$1 trillion.

On the same day, Tencent's market value was about HK$3.94 trillion, Alibaba's was about HK$1.97 trillion, Xiaomi's was about HK$612.9 billion, Meituan's was about HK$444.5 billion, JD.com's was about HK$302.9 billion, Ctrip's was HK$229.3 billion, and Kuaishou's was about HK$196.9 billion.

A company that has been established for less than seven years, has not yet turned a profit, and had an annual revenue of only over HK$700 million in the past year has a market value higher than that of a large number of China's most profitable internet companies.

The capital market is using real money to express a judgment: the gold mines marked on the old map are depreciating; the outline of the new continent is emerging, and the pricing power has begun to shift.

01

Under the Shadow of Trillion - Dollar Market Value

Most market interpretations of Zhipu AI stay at the technical level. The GLM series has achieved good results in evaluations. Its open - source strategy has attracted developers, and the number of enterprise customers is increasing. The official response on X about the technology competition over the weekend has further fueled this wave of market trends.

The waves are lively. But it has never been the waves that determine the direction of the tide.

Tencent's net profit last year exceeded HK$200 billion, yet its price - earnings ratio is less than 15 times. Alibaba's market value is even lower than the sum of the value of its stake in Ant Group and the value of its core e - commerce business. The capital market is re - pricing in a cold - blooded way: the existing assets of internet giants are depreciating.

Meanwhile, a large - model native company with an annual revenue of HK$700 million and still in the red has received a valuation of HK$1 trillion.

What the capital is clearly betting on is not the intrinsic value, but the disruptive replacement of the business model: the replacement ability of large models - to replace search, matchmaking, content production, and code writing - will systematically erode the most core profit pools of internet giants in the next decade.

Of course, many people are also saying that Zhipu AI's trillion - dollar market value is just "greater fool theory", an impulsive pricing by the market under the technology boom. This kind of doubt is not entirely unfounded. No matter how one dissects the cash flow of a still - loss - making company, it is difficult to justify a trillion - dollar valuation.

However, the label of "greater fool theory" may precisely obscure the paradigm shift in the valuation logic: when the moats of internet giants develop structural cracks due to large models, the transfer of market share has a high - probability path. The market is no longer pricing Zhipu AI's present, but the expected timeline of large models devouring the old profit pools.

The giants have seen this trend but find themselves unable to move.

Tencent's WeChat ecosystem relies on advertising revenue. When users no longer scroll through Moments or read official accounts, but directly ask AI assistants for information, the commercial foundation of WeChat begins to shake.

Alibaba's e - commerce empire is built on the chain of "search - compare prices - place an order". When AI assistants directly help users select products and compare prices, the traffic entrances of Taobao and Tmall are rendered ineffective.

Even ByteDance, which has been the most aggressive in AI investment, faces a similar logical challenge: Douyin retains users through its recommendation algorithm. When AI can directly generate the content users want to see, the moat of the recommendation algorithm is bypassed.

Large models have quietly maneuvered behind the three broadest moats of the internet era: you know where the enemy is coming from, but your city walls are built in the other direction.

02

Identity Crisis

At this moment, internet giants are facing a deeper dilemma: if large models make the existing business models meaningless, then will they still be the same as before if they become (or fail to become) winners in the large - model field?

Tencent's core asset has never been WeChat advertising, but "connection" - WeChat connects people with people, official accounts connect people with content, and mini - programs connect people with services. Advertising, payment, and game distribution are all by - products of connection.

In the era of large models, the interaction logic is shifting from "connection" to "agency". Users no longer open individual mini - programs or browse individual official accounts, but directly tell AI assistants "book me a flight ticket" or "write me a weekly report". AI assistants directly call services, generate content, and complete tasks, and the connection is compressed into the background.

There is an additional intermediate layer between users and Tencent. The more powerful this intermediate layer is, the farther Tencent is from users. The more users get used to obtaining everything through AI assistants, the lower the strategic value of WeChat as a "connector" becomes. What Tencent is facing is the loosening of the meaning of the company's existence. When your core mission is connection, and the era tells you that connection is no longer scarce, who are you?

Alibaba is facing a race between two curves. One is Alibaba Cloud. Large - model training and inference require a huge amount of computing power. The stronger Tongyi Qianwen is, the greater the API call volume will be, and the more certain the revenue growth of cloud infrastructure will be. The other is the advertising revenue of Taobao and Tmall. When AI assistants directly help users select products and compare prices, information asymmetry is reduced, and merchants' willingness to buy traffic will weaken accordingly.

This is not something that will happen tomorrow, but the direction is certain. Alibaba is standing on these two curves simultaneously, and their trends are not related. How fast Alibaba Cloud can grow and whether it can support a new growth pole before the e - commerce advertising revenue hits the ceiling is the real test of this transformation.

ByteDance's dilemma is actually more hidden. Although it seems to be all - in on AI, it inevitably struggles. Douyin's recommendation algorithm is based on the precise capture of user behavior. What users watch, how long they stay, and who they like, every action feeds the algorithm. The recommendation logic of large models is shifting from "behavioral data" to "semantic understanding". It doesn't need you to watch a thousand videos to know what you like; you just need to say "I'm in a bad mood today". The barrier of behavioral data may be worthless in the face of semantic understanding.

In addition, the Doubao series has performed well in evaluations, but ByteDance cannot abandon its most profitable recommended advertising business. Large models are still limited to the role of "auxiliary recommendation" within ByteDance and cannot touch the core business logic. The reason is obvious. This is not a matter of courage, but is determined by the company's DNA. A company that believes in "behavioral matching" finds it difficult to switch its soul to "semantic understanding".

The three giants, with three kinds of confusion, share the same underlying logic: the core asset in the era of large models is "intelligence", while the core asset in the internet era is "data". Data can be accumulated, but intelligence is universal. No matter how much data is accumulated, a more powerful foundation model can surpass it in an instant.

This kind of paradigm conflict has been repeated in business history. Kodak invented the digital camera but locked it in the safe because digital would kill film. Nokia made the touch - screen phone earliest but was reluctant to give up the physical keyboard.

Every time there is a technological paradigm shift, the giants of the old era will painfully waver between "protecting existing profits" and "embracing the new paradigm".

The cruelty of the large - model era lies in the fact that it won't give you ten years to transform slowly. It took less than three years from GPT - 3 to the trillion - parameter model, while Kodak struggled for at least twenty years from the invention of the digital camera to its bankruptcy. The window is narrowing at a visible speed, and the steering wheel is still locked in the profit logic of the old era.

03

Those Without Burdens

Zhipu AI can receive a valuation of HK$1 trillion precisely because it doesn't have these burdens. It doesn't need to protect advertising revenue, worry about the interests of e - commerce platforms, or maintain the moat of the recommendation algorithm. It can unreservedly bet all its resources on improving model capabilities and building a developer ecosystem.

The winners in the internet era - Tencent, Alibaba, and ByteDance - first gathered a large number of C - end users and then monetized through advertising, e - commerce, and games. Technological capabilities serve the business model.

Large - model native players like Zhipu AI take a different path: first establish a generational advantage in model capabilities, attract developers, build an ecosystem, and then look for commercialization paths. OpenAI and Anthropic are no exception. They obtained sky - high valuations based on technological leadership before a clear business model emerged and then gradually established a commercialization closed - loop through enterprise - level services and developer payments.

Whether this path can succeed in China remains unanswered for now.

A company with an annual revenue of over HK$700 million and still in a large - scale loss stage needs to prove at least three things in the next few years to support a market value of HK$1 trillion: the model capabilities can continuously approach the global first - tier; the revenue from API and enterprise services can increase rapidly; high R & D investment can ultimately lead to higher gross margins and stronger repurchase. Otherwise, it will be difficult to maintain today's valuation just by telling stories.

But the market has already given a valuation of HK$1 trillion. This price at least shows one thing: Investors believe that the business model in the era of large models will not naturally grow from the profit pools of the old era. It requires a company without historical burdens to cultivate it on new soil.

04

Conclusion

The identity crisis of Chinese internet companies is ultimately due to the invalidation of the coordinate system. The capital market was the first to sense this.

The crack between the two sets of coordinates is the collective strategic confusion of Chinese internet companies themselves. Everything measured by the old coordinates is depreciating. The new coordinates are in the hands of a few people. Most people are still standing on the shore of the old continent, looking at the masts in the distance, not knowing whether they should set sail.

Zhipu AI's trillion - dollar valuation is not the victory of one company, but the silence of countless companies. An old map is being folded up, and a new map is just being unfolded. The giants of the old era are stuck on the shore, watching the masts of others disappear over the horizon. No one can tell them whether what awaits them after setting sail is a new continent or deeper seas.

This article is written based on publicly available information and is for information exchange purposes only. It does not constitute any investment advice.

This article is from the WeChat official account "Jinduan" (ID: jinduan006), written by Mu Yang and published by 36Kr with authorization.